What if AI is not the right answer for our problem?

We will say so, and it happens regularly. Opportunity mapping exists to answer that question honestly, including the answer that a workflow is better served by a simpler fix, usually automation, better data, or removing a step.

How do you handle accuracy and risk?

Error tolerance is a design input. Low-tolerance decisions get low autonomy, cited evidence and mandatory review. High-tolerance, high-volume, reversible work can take more automation with sampling. The interaction design follows the risk profile rather than the model's capability.

Specialism / AI Product Design Agency

The model is rarely the reason the AI product failed.

We have yet to see an AI adoption failure caused primarily by model quality. The failures are in the experience and in the workflow the capability was dropped into.

Start a conversationAI & Product Design in full

Who this page is for

For teams building AI products who need design and workflow expertise, not modelling.

Related capability

Product design
SaaS design
Mobile app design
Enterprise UX
AI product strategy
AI experience design
AI UX

What we do and what we do not

We do not build models. We work with your data science team, your engineering team or your vendor. Our work is the product, experience and business design around the capability.

That includes the autonomy model, the interaction patterns, how evidence is disclosed, how correction works, what the interface does while the system is uncertain, and how the surrounding process has to change for the value to land.

Evidence-first interfaces

Professional users cannot act on conclusions they cannot verify, particularly when they carry the liability. An assistant that states a conclusion without provenance saves nobody any time, because the expert reads the source material anyway.

We design findings that cite their sources, open them in place, and express uncertainty in language rather than percentages. Showing where each answer came from does more than reassure people: it lets them spot a wrong answer quickly.

The workflow is where the value is

An AI capability inserted into a process designed for unassisted human work produces no measurable benefit, because the gain is reabsorbed by handovers, approvals and rework that no longer need to exist.

The highest-value part of most AI engagements is not the interface. It is the six weeks spent with risk, compliance and operations redesigning what happens around the capability.

Signals

When this work is worth doing

If two or more of these are true, the conversation is probably overdue.

The pilot performed and adoption did not followAlmost always an experience or workflow failure, not a model failure.
There is an AI mandate and no thesisA list of features with no connection to a cost line or a customer problem.
Users do not trust the outputUsually a provenance and correction problem rather than an accuracy problem.
Nobody has changed the processWithout process change, efficiency gains are absorbed before they are measured.

How we scope it

Three ways this usually runs

Durations are typical rather than fixed. We scope after the first conversation, not before.

AI opportunity mapping
4 to 6 weeks
Where AI would genuinely change your economics, qualified against data, risk and readiness.
You receive
- Workflow & decision map
- Qualified opportunity portfolio
- Readiness assessment
AI experience design & prototype
8 to 12 weeks
Autonomy model, interaction design and a working prototype tested against real failure cases.
You receive
- Autonomy & control model
- Working prototype
- Validation findings
Productisation
12 to 20 weeks
Production design, workflow redesign, evaluation and monitoring, and AI pattern library.
You receive
- Production design
- Target-state workflow
- AI pattern library

AI Product Design Agency

Questions we hear before we start

No. We design the product and experience around them, and redesign the workflow so the capability produces measurable value. We work with whoever builds the model. Your team or a vendor.

The thinking behind this work

All insights

AI28 August 20269 min read

Why AI products need a different UX framework

Conventional interaction design assumes the system does what it is told. AI products break that assumption, and most of the patterns we rely on break with it.

Read the thinking